www.sciencemag.org SCIENCE VOL 335 30 MARCH 2012 1579

PeRsPeCtiVes

findings do not rule out a role for mTORC2 in aging. What Lamming et al. do achieve is a clear dissociation of impaired glucose homeostasis and enhanced longevity. Indeed, mtor+/− mlst8+/− mice have normal glucose homeostasis, indicating that delayed aging derives from some other consequence(s) of mTORC1 inhibition.

Considerable effort is being directed to identify new and better mTOR inhibitors. The results from Lamming et al. lead to some early predictions about the ability of such compounds to modulate longevity. For instance, high-specificity inhibitors that tar-

get the active site of mTOR, such as PP242 and Torin, have recently been developed (14). Although blocking both mTORC1 and mTORC2 function may be ideal for oncol-ogy and other conditions, these inhibitors may be less effective than rapamycin for increasing longevity and preventing age-associated diseases. The study by Lamming et al. may point the way toward such thera-peutic approaches.

References and Notes 1. R. M. Anderson, R. Weindruch, Trends Endocrinol. Metab.

21, 134 (2010). 2. B. K. Kennedy, J. Intern. Med. 263, 142 (2008).

3. D. E. Harrison et al., Nature 460, 392 (2009).

4. D. W. Lamming et al., Science 335, XXX (2012). 5. M. N. Stanfel, L. S. Shamieh, M. Kaeberlein, B. K. Ken-

nedy, Biochim. Biophys. Acta 1790, 1067 (2009). 6. V. P. Houde et al., Diabetes 59, 1338 (2010). 7. E. Dazert, M. N. Hall, Curr. Opin. Cell Biol. 23, 744

(2011). 8. D. D. Sarbassov et al., Mol. Cell 22, 159 (2006). 9. C. Selman et al., Science 326, 140 (2009). 10. P. Kapahi et al., Curr. Biol. 14, 885 (2004). 11. T. Vellai et al., Nature 426, 620 (2003). 12. A. A. Soukas, E. A. Kane, C. E. Carr, J. A. Melo, G. Ruvkun,

Genes Dev. 23, 496 (2009). 13. D. A. Guertin et al., Dev. Cell 11, 859 (2006). 14. Q. Liu et al., Methods Mol. Biol. 821, 447 (2012).

10.1126/science.1221365

How Smart Is Your Home?computer science

Diane J. Cook

Technical advances are bringing intelligent homes that respond to residents’ needs and wishes within reach.

Individuals spend most of their time in their home or workplace; for many, these places are their sanctuaries. Over

the course of the 20th century, technologi-cal advances have helped to enhance the comfort and shelter provided by our homes. Insights gained from capturing and mod-eling behavior in these places may be use-ful in making our environments more intel-ligent and responsive to our needs. Recent advances are bringing such “ambient intel-ligence” in the home closer to reality.

Since the miniaturization of micropro-cessors, computing power has been embed-ded in familiar objects such as home appli-ances and mobile devices; it is gradually pervading almost every level of society. Ambient intelligence extends the notion of computing to provide customized, auto-mated support that is so gracefully inte-grated with our lives that it “disappears” (1). In the home, the idea is that computer soft-ware playing the role of an intelligent agent perceives the state of the physical environ-ment and residents using sensors, reasons about this state using artificial intelligence techniques, and then takes actions to achieve specified goals, such as maximizing comfort of the residents, minimizing the consump-tion of resources, and maintaining the health and safety of the home and residents.

During perception, sensors embedded in the home generate readings while residents perform their daily routines (see the figure). The sensor readings are collected by a com-

puter network and stored in a database that the intelligent agent uses to generate use-ful knowledge such as patterns, predictions, and trends. On the basis of this informa-tion, the intelligent agent selects an action and stores this selection in the database. The action is transmitted through the network to the physical components that execute the command. The action changes the state of the environment, triggering a new percep-tion/action cycle.

Filling a home with sensors and control-ling devices by a computer is not only possi-ble, but it is simple and commonly found in homes today. Sensors are available off-the-shelf that localize movement in the home, provide readings for ambient light and tem-perature levels, and monitor usage of doors, phones, water, and appliances. Tiny, inex-pensive sensors can be attached to objects to not only register their presence but also record histories of recent interactions. Such smart objects can harvest their own energy, and recent standards facilitate vendor-inde-pendent plug-and-play sensor design and modeling (2).

Proliferation of sensors in the home results in large amounts of raw data that must be analyzed to extract relevant infor-mation. Most smart-home data from envi-ronmental sensors such as infrared motion sensors and magnetic door sensors can be processed with a small computer. Once data are gathered from wearable sensors and smart phones (largely accelerometers and gyroscopes; sometimes adding camera, microphone, and physiological data), the amount of data may get too large to handle

on a single computer, and cloud computing is appropriate. Cloud computing is also use-ful if data are collected for an entire commu-nity of smart homes to analyze community-wide trends and behaviors.

Currently, most users write rules by hand to interpret sensor data and to control devices. For example, home owners install-ing home automation equipment must write their own rules for when their lights turn on and off. Artificial intelligence (AI) plays a pivotal role in automating this process. AI and data-mining technologies seek useful information on resident behavior and the state of the home. Computer algorithms have been designed to predict and identify activities performed in the home and to rec-ognize emotions, body mannerisms, and gestures (3, 4). These capabilities, as well as the abilities to recognize activities, identify trends, make assessments, and take action, are becoming more available and robust, but are not commonly found in actual homes.

The goal of much current research in ambient intelligence is to enable devices to interact with their peers and the network-ing infrastructure without explicit human control. The intelligent home must also be imbued with an awareness of the resident context (location, preferences, activities), physical context (lighting, temperature, house design), and time context (hour of day, day of week, season, year). Providing this type of context-aware reasoning makes it possible to design environments that pro-vide, for example, customized lighting and temperature settings based on learned resi-dent preferences; information retrieval and

Washington State University, Pullman, WA 99164–2752, USA. E-mail: cook@eecs.wsu.edu

EMBARGOED UNTIL 2:00 PM US ET THURSDAY, 29 MARCH 2012

30 MARCH 2012 VOL 335 SCIENCE www.sciencemag.org 1580

PeRsPeCtiVes

displays that follow residents from computer screen to wall to tabletop; and automated reminders to residents of important tasks they need to perform (5). As AI research progresses, these services will increasingly be provided on the basis of detected behav-ioral patterns, rather than rules provided by the user, as is currently the case.

Because ambient intelligence seeks to be both unobtrusive and ubiquitous, it can affect almost every aspect of human life without drawing attention to itself. Three applications of ambient intelligence in the home have received particular attention: health monitoring, energy efficiency, and social computing.

Ambient intelligent homes offer tech-nologies for automated health monitoring, assessment, and intervention. For example, researchers have used capabilities found in ambient intelligent homes to find a link between changes in mobility patterns and the onset of symptoms of dementia (6). Motion sensors were placed throughout the home, and total space covered throughout the day as well as walking speed were esti-mated for individuals over multiple years. Changes in these parameters correlated with early symptoms of dementia. Other researchers have used ambient intelligence technologies to perform early-childhood screening for autism (7). Building upon con-text awareness, smart homes provide well-timed prompts to remind individuals to per-form familiar routines and to initiate new health behaviors (8).

Monitoring energy consumption in the home is also increasingly important, because energy consumption is rising at a higher rate than population growth, and households are responsible for over 40% of total energy use in most countries. Ambient intelligent homes facilitate smart energy management through nonintrusive load identification and estimation, as well as minimally intrusive load shedding and automation for energy efficiency (9).

Another aspect of daily life is social interaction. Social signals have long been recognized as important for establishing relationships, but only with the introduction of sensed environments have researchers become able to monitor and measure these signals (10). Building upon ambient intelli-gence technologies, we can look at social-ization within the home (such as entertain-ing guests, interacting with residents, or making phone calls) and examine the cor-relation between socialization parameters and productivity, behavioral patterns, and health. These results will help researchers

not just to understand social interactions but also to design products and behavioral interventions that promote interaction, or to leverage existing social relationships in ways that influence interventions and pur-chasing choices.

The dream of ambient intelligent homes is, however, hampered by some formidable challenges. A primary concern is the need to consider possible implications for privacy and security. Many individuals are reluc-tant to introduce sensing technologies into their home, wary of leaving digital trails that

others can monitor and use to their advan-tage, such as to break in when the house is empty. To allay such concerns, researchers are investigating ways to define and pro-vide guarantees for levels of privacy and for the safety of the technologies. Similarly, smart homes need to ensure that the resi-dent retains the ultimate authority to reset the system and to impose constraints that prevent the home from taking undesired or harmful actions.

Another critical challenge for home-based ambient intelligence is to provide

4

17

17

17

1

16

16

15

14

13

12

12

11

109

8

7 7

7

6

555

55

5

55

5

4

3

3

3

3

2

17

17

17

1 AmI control 2 Light sensor 3 Windows and door control 4 Hvac control 5 Lighting control

6 Automatic pet feeder 7 Motorized drapes 8 Automatic watering 9 Mailbox sensor 10 Driveway sensor

11 Security system 12 Lawn moisture sensor 13 Face recognition sensor 14 Motion sensors 15 Door sensors

16 AmI interface with car 17 AmI interface with smart phone

Coming soon to your home? In an ambient intelligent home, sensors collect information about the envi-ronment and the residents. An “intelligent agent” uses this information to decide whether actions need to be taken to adjust e.g., temperature or lighting. C

RE

DIT

: aD

apT

ED

fR

om

ISTo

Ck

pho

To.C

om

EMBARGOED UNTIL 2:00 PM US ET THURSDAY, 29 MARCH 2012

www.sciencemag.org SCIENCE VOL 335 30 MARCH 2012 1581

PeRsPeCtiVes

The idea of ambi-ent intelligence implies an intrin-

sic link between indi-viduals and their envi-ronment, enabling indi-viduals to access and interact with computing artifacts in ways that are intuitive and do not dis-rupt everyday activities. Given the many different environments encoun-tered as part of everyday life—within the home (1) as well as beyond it—enabling such interaction is a formidable techno-logical challenge. The reward may be an envi-ronment that is safer, uses less energy, and responds to the needs of all individuals (see the figure). Recent advances in embedded systems, robotics, and sensor technology suggest that ambient intelligence may indeed be realized, particularly if crucial privacy and security concerns are addressed.

Three examples illustrate the poten-tial of ambient intelligence across differ-ent domains. First, information filtering and personalization in a museum context offer a practical and eminently achievable vision of ambient intelligence. Second, autonomous mobile robotic services are a development that challenges perception, trust, and accep-tance of ambient intelligence. Third, a sim-ple ambient intelligence smart-phone ser-vice shows how the quality of life of those afflicted with dementia can be improved.

Museums and art galleries comprise many exhibits, yet tend to offer broad gen-eralized information, usually in the form of visitor or audio guides. Ambient intelligence challenges this “one-size-fits-all” approach, envisaging a digital information space that enables the personalization of information to meet diverse user needs (2). This may include generating recommendations for visiting other exhibits, based on, perhaps, what the visitor has already seen and what their cultural interests are. Typically, a con-ventional smart phone is sufficient to enable this kind of interaction, acting as a lens into the information space. In sharing positional and personal information, multimedia con-tent may be personalized and presented via the visitor’s smart phone, resulting in a more satisfactory experience. However, a prereq-uisite is the deployment of a suitable ambi-ent intelligence technological infrastruc-

ture—using radio-frequency identif ication (RFID), for example—and the construc-tion of an information space

for all the exhibits. This latter issue is of prime importance, demanding that the cura-tor has a deep understanding of the exhibits and the visitors they hope to attract.

Ubiquitous robotics (3) represents an area of substantial potential for ambient intelligence in the longer term. Many of the technologies to successfully deploy suites of mobile robots exist. For example, the Dust-Bot project (4) has demonstrated robots collecting waste in an urban environment without direct human oversight, demanding obstacle (stationary and mobile) avoidance on the part of the robot, as well as interaction with householders. Computational intelli-gence, embedded both on the robot and in the environment, is essential for enabling this behavior. However, at present the cost remains excessive. Until this is addressed, robots cannot be deployed in a widespread fashion, and the key issues of human-robot interaction and social acceptance cannot be researched thoroughly.

Ultimately, ambient intelligence is about people and improving their quality of life. Consider the case of people with dementia, an issue that will have increasing implications for society in the coming years. Wandering is characteristic of many with dementia and is a major factor contributing to institutionaliza-C

RE

DIT

: mIC

ha

Il l

og

Ino

v/I

STo

Ck

pho

To.C

om

University College Dublin, Belfield, Dublin 4, Ireland. E-mail: michael.j.ogrady@ucd.ie

intelligent support seamlessly both inside and outside the home. With the introduc-tion of wearable sensors and smart phones, ambient intelligence is not limited to the home environment (11). Our ambient intel-ligent agent needs to seamlessly merge sup-port inside the home with that available in a mobile setting. Smart homes currently offer useful monitoring and automation with guid-ance from the resident. As the field matures,

we anticipate that these services will be pro-vided in a more independent manner.

References and Notes 1. M. Weiser, Sci. Am. 265, 94 (1991). 2. E. Y. Song, K. Lee, IEEE Instrum. Meas. Mag. 11, 11 (2008). 3. E. Kim, S. Helal, D. Cook, IEEE Pervasive Comput. 9, 48

(2010). 4. J. Alon, V. Athitsos, Q. Yuan, S. Sclaroff, IEEE Trans. Pat-

tern Anal. Mach. Intell. 13, 1685 (2008). 5. G. M. Youngblood, D. Cook, IEEE Trans. Syst. Man Cybern.

C Appl. Rev. 37, 561 (2007).

6. T. Buracchio, H. H. Dodge, D. Howieson, D. Wasserman,

J. Kaye, Arch. Neurol. 67, 980 (2010). 7. A. Dey, D. Estrin, IEEE Pervasive Comput. 10, 4 (2011). 8. B. Das, D. Cook, D. Proceedings of the 9th International

Conference on Smart Homes and Health Telematics (Springer, Berlin, Heidelberg, 2011), pp. 9 –16.

9. J. Paradiso, P. Dutta, H. Gellersen, E. Schooler, IEEE Pervasive Comput. 10, 11 (2011).

10. S. Pentland, Am. Sci. 98, 204 (2010). 11. M. O’Grady, G. O’Hare, Science 335, xxxx (2012).

10.1126/science.1217640

How Smart is Your City?From art galleries to our cities’ streets, ambient intelligence can help to make the human environment more responsive to the needs of individuals.

Michael O’Grady and Gregory O’Hare

A changing world. Through smart-phones and related devices, users have the world at their fingertips. But what if the city environment could respond automatically to indi-viduals’ needs and wishes?

computer science

EMBARGOED UNTIL 2:00 PM US ET THURSDAY, 29 MARCH 2012